The panopticon of Germany’s foreign trade, 1880–1913: New facts on the ﬁrst globalization

We present and analyze the panopticon of Germany’s foreign trade, with new data on all products, all trade partners, quantities, and values, at annual frequency, 1880–1913. Historical product categories are reclassiﬁed according to the Standard International Trade Classiﬁcation (SITC) to ensure comparability over time and across countries. Germany became increasingly specialized in manufacturing, in line with theories of comparative advantage. However, most trade growth occurred along the extensive margin, and 20–25 percent of trade was intra-industry trade, at ﬁve-digit SITC. Both facts suggest substantial within-sector heterogeneity. We discuss why this matters for our understanding of the ﬁrst globalization.


Introduction
"German trade statistics before 1906 are a booby trap."Lewis (1978, p. 26).By 1914, Germany had risen to the second largest exporter in the world, just behind the UK.Ever since, Germany has been a major player in international trade.A large literature explores the country's integration and role in the world economy.The merchandize trade dimension of Germany's globalization alone has been subject of research for more than a century. 1 In particular, researchers have been trying to understand how Germany could develop from a backward, predominantly agricultural economy around 1850 into the industrial core of the global economy by 1914.The dominant explanation is one of declining trade costs that facilitated a specialization along the lines of comparative advantage.According to this view, German trade expanded in line with advantages in physical and human capital at the expense of labor-and land-intensive sectors, notably textiles and agriculture.
However, research on the history of foreign trade for any country before 1945 had often to rely on aggregated data or focus on limited case studies.Still today, researchers struggle with imprecise, not fully comparable data of international trade.Using historical trade data comes with even bigger problems: product categories are entirely unstandardized, the degree of detail varies, and misreporting is widespread.The quote by Arthur Lewis above indicates that handling historical trade data from Germany is particularly difficult.Indeed, the German data are highly incomplete before 1880, they lack a harmonized product classification and the trade of Germany's two largest ports (Hamburg and Bremen) was excluded from official statistics before 1889.On the other hand, the data contained in original historical sources are comparatively granular and thus rather rich.They are a treasure not lifted until now.
In this paper, we present a new dataset that provides what we call the panopticon of Germany's foreign trade .We show that this new evidence challenges the traditional perspective on the first globalization in several ways.In the first part of this paper, we briefly describe our main sources and methods to construct the new dataset.We digitalized all available historical records from original sources, integrated Hamburg and Bremen into the German customs union (Zollverein) data,2 and reclassified all data according to the Standard International Trade Classification (SITC, Rev. 4).In the second part of the paper, we are thus able to describe and analyze Germany's integration into the world economy during the first globalization in a way that was impossible before.This connects our work to a growing literature that reexamines economic history with newly digitalized, much bigger datasets (Abramitzky 2015;Mitchener 2015).Moreover, the way in which the data are compiled allows direct comparison to modern trade data and thus lays the foundation for a disaggregated long-term series on Germany's international trade.Doing so, we contribute to some major revisions the historiography of international trade is currently experiencing (e.g., Meissner 2015, Huberman and Oosterlinck 2017, Becuwe et al. 2018, Becuwe et al. 2021).Specifically, the complete coverage of our data at the product-country level allows us to decompose total trade growth into contributions from the intensive and extensive margin.
Our evidence challenges the existing narrative for the first globalization, based on variants of the Heckscher-Ohlin-Vanek model (O'Rourke and Williamson 1999).We show that virtually all growth in German exports, 1880-1913, and about 80 percent of total growth in German imports took place along the extensive margin of trade.This strongly suggests that within-sector heterogeneity was crucial already during the first globalization, possibly due to monopolistic competition with fixed costs and economies of scale.Second, we document that intra-industry trade was important not only for both exports and imports but also for manufacturing and non-manufacturing trade.Again, this points to a role for heterogeneity within sectors, but even within products (which we define as a five-digit SITC category).Third, while running an overall trade deficit for each year in our data , Germany achieved a growing trade surplus with rich and neighboring trade partners.Most of this surplus was due to growing exports in manufacturing sectors of the second industrial revolution, which dominate German exports until today: chemicals, machinery, and transport equipment.These latter developments are in line with models based on comparative advantage.Overall, this suggests that we need to revise our view on the first globalization.Germany did broadly specialize along comparative advantage, but both imports and export dynamics suggest a large role for within-sector heterogeneity.This matters because it affects our interpretations about the gains from trade, their distribution, and industry reallocations before 1914.
We proceed as follows.Section 2 discusses the major challenges with German trade statistics, our sources, and how we constructed the new dataset.In Section 3, we use the data to describe German trade growth along four basic dimensions.In Section 4, we decompose trade growth in terms of intensive and extensive margin trade and show the role of intraindustry trade.In Section 5, we analyze the patterns of specialization and diversification over time.In Section 6, we summarize our findings, place them into the context of the recent literature, and discuss some broader implications for the first globalization.

Constructing German trade data before 1914
Our main data sources are the import and export tables for special trade in various volumes of the Waarenverkehr des Deutschen Zollgebiets mit dem Auslande, published by the Imperial Statistics Office (Kaiserliches Statistisches Amt). 3 We focus on special trade ("Spezialhandel"), which differs from general trade by excluding transit ("direkte Durchfuhr"). 4As we explain below, for the years 1880-1888, we augment this data for the Zollverein with data for Hamburg from the Tabellarische Übersichten des Hamburgischen Handels, published by Handelsstatistisches Bureau and Bremen from the Statistik des Schiffs-und Waarenverkehrs in the Jahrbücher für Bremische Statistik published by Bureau für Bremische Statistik.We recompile, rearrange, and reclassify these data to arrive at a disaggregated but harmonized trade dataset ready for economic analysis.The major challenges for work with German trade statistics before 1914 relate to the incompleteness of data before 1880, the manifold changes in product categories over time, and the fact that Germany's two largest ports were treated as foreign trade partners in the statistics, because they stayed outside the German customs union until the end of 1888 (Torp 2005).We show that despite these problems, a complete disaggregated trade dataset is obtainable from 1880 onwards.
German trade data before 1880 is patchy.5Before the German Empire was founded, the only German source on Germany's international trade are records on tariff collection by the different member states of the German Customs Union (Zollverein), the so-called Kommerzialnachweisungen (commercial certificates).These records are, however, very crude.They capture only goods subject to tariffs, refer only to observed quantities not values, and countries of origin are largely missing (only the entry point to the Zollverein is recorded). 6fter the foundation of the German Empire in 1871, the first unified trade statistics and, generally, more encompassing official statistics emerged.7However, under a largely freetrade regime, there was little need to specify the observation of external trade, and only trade subject to tariffs was systematically recorded.Related, the early German trade data suffers from misdeclarations.Instead of the actual trade partner, the data report the country that hosts the last port of entry.Due to the introduction of protectionist tariffs in 1879, Germany started to collect data on all trade flows, covering all goods and trade partners from 1880 onwards.Overall, given that statistical information on trade before 1880 is incomplete or altogether missing, we start our data reconstruction in 1880.
A major challenge for the construction of time series is that the number of product categories in the trade statistics changed and grew repeatedly over the years.The number of products increased, either because actually new products were traded or because the statistical authorities changed the degree of detail to which product categories were differentiated.The latter typically happened because some new type of product became quantitatively more important as an import or export.An important impulse for changes in the classification system was new tariff laws, which required more detailed statistical information, notably the tariff reforms of 1879 and 1906.As a result, the number of categories quadrupled between 1872 and 1906, from 457 import positions to 2,030 and from 403 export items to 1,879 (see also Torp 2005, p. 55).Previous authors who attempted to make use of the granularity of the German trade data relied on rather ad-hoc, typically non-hierarchical reclassifications or they picked individual goods that they then compared.Therefore, two most detailed previous accounts of Höpfner (1992) and Jasper (1996) fail to provide an adequate image of the product structure of Germany's trade growth during the first globalization.Our approach is to use the SITC rev 4 because it allows comparison to other countries and more recent trade data.We refer to the reader to Hungerland and Altmeppen (2021) for details in the reclassification of goods, who discuss alternative approaches and conclude that SITC rev. 4 is best suited for historical data.
Moreover, until 1888, some parts of the German Empire stayed outside of the German customs union (Zollverein).The port cities Cuxhaven and Geestemünde and the important Hanseatic cities Hamburg and Bremen joined the Zollverein only in 1888 (the 'Zollanschluss'). 8These trading hubs kept own trade records-at varying levels of detailwhile the federal tariff area statistics treated these cities as foreign countries.The Hanseatic port cities were, however, very relevant to Germany's external trade: they commanded roughly a fifth of Germany's trade.As a result, the geographical (and commodity) structure of Germany's trade was distorted, as Buchheim (1982, p.24) notes: countries that were the major trading partners of the Hanseatic cities are underrepresented in the statistics of the German tariff area while countries that traded relatively less with the Hanseatic cities are overrepresented. 9To deal with this, we first reclassify all data from the German Zollverein, Hamburg, and Bremen based on the SITC system.Next, we combine the three datasets extending the quota method suggested by Buchheim (1982).For further details on the method, underlying assumptions and a discussion of the quota approach see Appendix A.2. Applying the quota method to our data yields a new regional structure of Germany's trade: trade with Europe decreases while trade with all other regions increases, both in absolute and in relative terms (see Table 1).Overall, the share of exports to Europe decrease by 14 percentage points (from 90 to 76 percent), while the share of imports decreases by 9 percentage points (from 91 to 82 percent).Once we differentiate further between Eastern and Western Europe, trade with the former increases, while trade with the latter decreases.The changes on the country level are even larger.For example, imports from Brazil increase by 614 percent in 1888 while exports increase by 263 percent.Other large changes (say more than 200 percent) occur for Asia, China, Japan, as well as Peru and Portugal.
Overall, this leaves us with a new dataset on German Foreign trade, covering all years 1880-1913, including the major ports Hamburg and Bremen for imports, exports, quantities, and values, with some qualifications on the latter. 108 Entry to the tariff area statistics was only fully consummated by 1906, except for some small areas in Baden (Badische Zollausschlüsse) which remained outside the tariff area. 9Two examples: In 1885, 48 percent of exports from Hamburg went to the UK, but only 6 percent of exports from the German tariff area.In 1889, after Bremen's and Hamburg's entries, exports to the United States grew allegedly by 67 percent. 10Similar to the approach in other countries (e.g., France, see Becuwe et al. 2018, p. 436), both imports and exports were valued by multiplying quantities for each product category with average prices that were defined annually by an expert commission.While these prices were often not country specific, this starts to change from 1889 onwards.After 1906, export values were (in most cases) based on declared values while imports continued to be valued by a commission until 1928 (see Hoffmann 1965, pp. 525-532).

The basic dimensions of Germany's trade growth
Let us start exploring the new data.Conceptually, we can distinguish between five dimensions of a country's foreign trade.Trade can vary in terms of quantities, prices (or unit values), the direction of trade, products traded, and trade partners.In this paper, we will focus on values of imports and exports at the level of trade partners and products.We start with an aggregate view on trade values and the balance of trade.

Imports and export values and the balance of trade
Figures 1A and 1B display the background for all what follows below: the development of aggregate exports and imports and the trade balance between 1880 and 1913 (1A), and Germany's exports and imports relative to the UK (1B).Germany's foreign trade more than tripled in the period we look at.During this period, Germany underwent its "second industrialization" with industrial growth especially in machinery, chemicals, and electrotechnical goods.Figure 1A suggests that substantial trade growth in both imports and exports accompanied this industrial growth.Nonetheless, the German Empire was running a persistent trade deficit throughout the period.Moreover, figure 1B shows that Germany was catching up to the biggest trader in the world before 1914, the UK.In 1880, the value of goods exported by the German Empire was about 47 percent of the UK's exports; by 1913, this figure had reached 95 percent.Similarly, the value of German imports increased from 42 (1880) to 83 percent (1913).While this evidence on the growth of German trade is not really new, our data allow us to explore the underlying changes in the structure of trade.We first discuss evidence on the country space, which is the number and importance of trade partners over time, followed by a discussion of the product space.

Country space
According to our data, the German Empire traded between 1880 and 1913 with up to 88 trade partners.Within this country space, the number of active trade relations was changing and generally growing over time.On average, Germany had positive trade flows with 66 of them.Table 2 and figure 2 show how the number of active trade partners developed over time, where we distinguish by our source for the years 1880-1888.We see how the inclusion of Bremen and Hamburg into the Zollverein statistics in 1889 led to a sudden (and artificial) increase in the number of Zollverein trade partners.Throughout the following discussion, we use our new data, where we merged Bremen and Hamburg to the Zollverein from 1880 onwards.
The changes in the number of trade partners after 1888 are due to an extensification of trade, where new markets were entered and thus only started to be documented in the records once trade flows with them turned non-zero.Another reason might have been the intensification of trade and, in turn, a more precise recording of existing trade partners.In our discussion of growth along the extensive margin below, we will attempt to distinguish between these two possibilities. 11 Figure 3A shows the evolution of trade shares by continents.While Europe always made up more than 60 percent of total trade, its share declined over time.On the other hand, trade  with non-Europe grew.This points to growing regional diversification in Germany's foreign trade, and it suggests that comparative advantage might have varied regionally.Figure 3B displays the regional trade balances, shown as averages over the entire period.Trade with the Americas was most in deficit, while trade with not only Europe but also Africa and Oceania much was less so.The figures show that imports (figure 4) were geographically less diversified than exports (figure 5), a common finding for European trade relationships (e.g., Federico and Wolf 2012).We note that Europe dominates for both imports and exports, but much more for the latter.For imports, the most important trade partner was the United States, followed by the Russian Empire.The largest trade partners for German exports were the UK and the Austria-Hungary.

Product space
The most valuable feature of our data is its enormous detail on the product space.We observe trade at the product level, which we define here as the lowest (namely five-digit) level of the SITC system.With the data classified according to the SITC, we can aggregate the data from the product level up to sectors.Moreover, we are able to apply other classifications compatible with SITC to our data and make comparisons to modern data for Germany and other countries available in SITC format.Figure 6 gives the basic characteristics of the product space.As in figure 2, we distinguish by our source for the years 1880 to 1888.We see again how the inclusion of Bremen and Hamburg into the Zollverein statistics in 1889 led to a sudden (and artificial) increase, this time in the number of listed products.
In figure 6A, we show that the number of SITC items observed in the data grows over time, corresponding to a growing number of original product categories in our source data. 12The latter, however, varied more wildly from year to year as Hungerland and Altmeppen (2021) discuss in great detail.But the growing number of items is also a function of technological progress and political economy that occurs in modern trade data as well: innovations generate new goods and lobbies successfully initiate the creation of special product categories that would enjoy a special status in a country's tariff scheme.Figure 6B gives an indication of the  precision with what we observe product-level trade.It shows the average (trade-weighted) length of an SITC code we were able to assign.Most trade flows are captured at the four-or even five-digit level.This makes our data one of the most granular datasets on trade available for the period before World War I.Moreover, the precision grows over time and remains high after the Bülow tariff of 1902, which was implemented in 1906. 13The figure also shows that Hanseatic data are less precise than the Zollverein data.Figures 7A and 7B describe the evolution of trade shares at the level of broad (one-digit) sectors.The first impression is that endowment effects (such as a relative abundance of physical and human capital) appear to structure Germany's foreign trade, as we would expect for example from a Heckscher-Ohlin logic: imports (figure 7A) are dominated by SITC sectors 0 to 3, that is, food, beverages, and raw materials.These items make roughly three quarters of Germany's imports.Exports (figure 7B), on the other hand, are dominated by manufactured products, that is, SITC sectors 5 to 8. Hence, their share grew over time.The data thus broadly confirm that Germany was increasingly specializing in exporting manufactured goods, while importing rather raw or less processed or more primary goods.
Table 3 reports the growth rates of sectoral trade from 1880 to 1913.The data show substantial variation in sectoral growth in both directions of trade and clearly modify the evidence from figure 7. Trade grew in nearly all sectors (except for imports of beverages and tobacco and animal and vegetable oils).The biggest growth rate is that of imports of commodities are not elsewhere specified, which however makes only a small part of the overall imports, as indicated by figure 7A.More tellingly, exports of machinery and transport equipment grew by a whopping 733 percent.Moreover, we see that imports of manufacturing goods grew very substantially and that the export of mineral fuels and lubricants increased, probably as a result of the growing role of the chemical industry, which is not solely covered by SITC sector 5. Hence, there is evidence for trade dynamics-notably in the manufacturing sector-that are not easily explained by neoclassical trade models.
Figures 8 and 9 display the product space in terms of the values of imports and exports in the year 1913.Again, given the hierarchical character of our data, we can use sunburst diagrams.They show the total of trade organized by SITC one-digit sectors on the inner circles and trade by SITC two-digit sectors on the outer circles.In this way, a more nuanced image emerges.Imports (figure 8) are generally more concentrated, with larger trade flows distributed over a smaller number of goods.Some goods such as cotton, hides, wheat, or copper stand out in that they command big chunks of the import volume.Exports (figure 9) are more diversified in terms of product groups, similar to the evidence on trade partners.Individual export goods generally command a smaller share in total trade than specific import goods.What remains striking, however, is that Germany also exported a range of goods we would not necessarily expect to be exported if we would follow a simple Heckscher-Ohlin prediction, such as coal or grains.

Varieties and margins decomposition
Our new data allow us to show that not only the value of aggregate exports and imports grew but also the number of trade partners and traded products changed very substantially over time.The total value of German trade grew from around seven billion mark (1880) to 21 billion mark (1913).Over the same period, the number of trade partners grew from We distinguish here between available and active varieties.Let us define available varieties as the number of observable country-product combinations given by all product classes and countries listed in the trade statistics in a given year.In 1880, we actually observe 245 different (SITC five-digit) products being imported and 248 being exported, out of a total 334 different product classes listed.Hence, many products must have been both imported and exported.Together with the fact that imports and exports are listed with the same 34 trade partners in 1880, we can calculate a total of 334 × 34 = 11, 356 available five-digit varieties as of 1880. Figure 10 shows the development of available varieties over time (right axis, shades show cumulative numbers) and the share of active varieties out of the number of available varieties at that time (left axis).We repeat this for higher levels of aggregation, namely fourand three-digit SITC groups.We see that the number of available varieties increased over time.Moreover, at any point in time, we observe that only a small but growing share of available varieties was actively traded.This holds for different levels of aggregation.Together, this evidence suggests to us that the growth in traded varieties is not just a statistical artifact due to changes in the classification system.Next, consider the evidence in table 4. If we consider only varieties listed in 1880 as available, the share of active in imports in 1880 was 9 percent, in exports 13 percent.Instead, if we assume that all countries and products listed in 1913 were in principle already available back in 1880 (but not yet listed), the share of active varieties in German imports in 1880 was just 1 percent, in exports 2 percent.Conversely, if we condition on the varieties listed in 1880, we see that for imports only 5 percent of those were active in 1913 and 9 percent for exports.Hence, while the total number of varieties increased over time very substantially, we also need to take into account that many varieties disappeared.A useful way to illustrate these dynamics is to decompose the total growth of trade into three margins, following Amiti and Freund (2010): first, growth along the intensive margin, where trade in existing varieties is expanding (more of the same); second, growth along the extensive margin, where new varieties enter; and third, growth along the extensive margin, where old varieties disappear.More formally, define a country-product pair k as a variety and denote the value of trade in variety k in year t as V kt .Then we have where the first term on the right hand side captures trade growth in existing varieties K t between t − 1 and t, the second growth in new varieties K N , and the third the disappearance of varieties K D . Figure 11 shows the result of this exercise.
We see that the total increase in imports (by 295 percent) can be decomposed into growth along the intensive margin (64 percent), growth along the extensive margin (entry) of 256 percent, and decline along the extensive margin (death) of −24 percent.For exports, a total of 261 percent growth is exclusively due to growth along the extensive margin, with −4 percent (intensive), +293 percent (extensive, entry), and −28 percent (extensive, death).This dramatic growth along the extensive margin reflects the diversification in traded products and the growing penetration of foreign markets.Books were traded to the United States, furs to Romania, toys to the UK, and silk to France.The largest trade partners showed also the biggest increase in traded varieties, notably the UK, Austria-Hungary, Russia, France, and the Netherlands.This evidence corresponds well with predictions of new trade models based on within-sector heterogeneity, such as Melitz (2003).In this model, which is the workhorse for many recent contributions to the theory and empirics of trade, firms are heterogeneous in terms of productivity and interact in differentiated product markets.A main implication is that a decline in trade costs will lead to an increase in the number of traded products, and hence growth along the extensive margin.But is it really plausible to find that all growth in exports took place along the extensive margin?We first show that the margin decomposition is different depending on sectors and trade partners, in line with an intuition building on new trade models in the wake of Melitz (2003).Next, we show that our results are unlikely to be driven by statistical artifacts such as changes in trade classification due to an intensification of trade.Figures 12A and 12B show our margin decomposition separately for manufacturing and non-manufacturing trade.As expected, growth along the extensive margins matters most for trade in manufacturing sectors, growth along the intensive margin plays a larger (but still very limited role) for nonmanufacturing sectors.
Next, figures 13A and 13B distinguish between Germany's trade with European and non-European partners.Intuitively, growth along the extensive margin should matter more for trade with the more similar European neighbors, while growth along the intensive margin should characterize trade with non-European partners, due to specialization effects (as suggested by Heckscher-Ohlin type models).
Again, we find our intuition confirmed.The importance of the extensive margin is impressive overall and in all these subsamples of the data.Even for the growth of imports from non-European trade partners, which would be paradigmatic for the well-known narrative of the first globalization, we find only one-third of total growth explained by the intensive margin.
An altogether different explanation for our findings would be that they are due to some type of statistical artifact.Let us first consider the role of Hamburg and Bremen.For the years 1880-1888, we can decompose trade growth separately for the main ports Hamburg and Bremen and the Zollverein.Looking separately at the Zollverein and Hamburg, we find again that nearly all growth in exports took place along the extensive margin.Instead, the (overall smaller) trade of Bremen was declining over this period and this shrinkage occurred mostly along the intensive margin.Hence, the procedure of merging the data for the years before  1889 is not driving our results.But what if most of our extensive margin growth stems from the mentioned intensification of trade, which led to a more detailed statistical description of existing trade flows?If so, the new products and trade partners observed in the data might have been present before but were not properly documented.How can we judge the relevance of this?First, the evidence in figure 10 suggests that also for given levels of classification there was a systematic increase in the number of traded varieties.Moreover, we suggest restricting our data to growth in those products and trade partners, which were observed already in 1880, hence the set of available varieties as of 1880 shown in table 4. Based on this, we can calculate Figure 15.Intra-industry trade.A: Manufacturing (SITC sectors 5 to 8).B: Non-Manufacturing (SITC sectors 0 to 4).In percent of total trade according to eq. 2. Note that we consider our data at different disaggregations, down to the product-level (SITC five-digit level).Source: Own calculations.a lower bound of growth along the extensive margin, where we exclude all trade flows that we observe after 1880 in new products and with new trade partners as a statistical artifact.Note that the total growth in terms of imports and exports is much lower.Also note that, in this exercise, growth along the intensive margin remains by definition unchanged (compare figure 11), as it always refers to growth in varieties, which existed already in 1880.Figure 14 shows the result.
For imports, we see that the intensive margin now accounts for just above 50 percent of all trade growth.However, for exports we still find that virtually all trade growth took place in terms of extending the range of trade partners (out of those observed in 1880) for products listed in 1880 or the range of products (out of those observed in 1880) to a wider range of trade partners.This almost certainly underestimates the extensive margin.Hence, we can safely conclude that the extensive margin played a crucial role for the overall growth of German imports 1880-1913 and totally dominated the dynamics of German exports 1880-1913.

Intra-industry trade
So far, we considered exports and imports separately, following a logic of differences along country characteristics (as in Armington 1969).In this section, we drop this and explore the extent to which the same products were both imported and exported by Germany.The sectoral growth rates of exports in sectors other than manufacturing in table 3 and the shares shown in figure 7 already suggested that this might have been an important phenomenon.We calculate indices of intra-industry trade (trade within a given product category) as in Grubel and Lloyd (1971): with k being the SITC level, t being the year, X being exports, and M denoting imports.The index gives the share of trade that takes place within a given sector. Figure 15 plots this index once for manufacturing (figure 15A) and once for non-manufacturing (figure 15B) and over all levels of aggregation. 14s we would expect, the more aggregated the flows are, the more intra-industry trade we find.Interestingly, the data reveal that roughly a quarter of trade happened within the most narrowly defined category, that is, the five-digit level.If we look at the one-digit level, we see that this share grows to about 50 percent of all trade.The figures also show that intra-industry trade was remarkably stable over time, for both manufacturing and non-manufacturing trade.Germany exported many non-manufacturing goods, such as coal or hides and skins.At the same time, Germany imported a wide (and growing) range of manufacturing goods before 1913.If Germany's foreign trade had followed simple comparative advantage alone, we would not see such intra-industry trade flows in the data.They make up, however, a nonnegligible share of Germany's trade value, particularly with Europe, as we show below.The fact that the level of intra-industry trade remained very stable over time is rather remarkable.Becuwe et al. (2021) calculate as Grubel-Lloyd index for French trade 1848-1872 at a level of disaggregation roughly similar to our three-digit level of SITC.They find that this index is strongly increasing to about 0.4, a level similar to what we find.Becuwe et al. (2021) explain this increase as the result of trade liberalization, notably in the wake of the Cobden-Chevalier treaty of 1860.Note that such an increase is at odds with a standard model of monopolistic competition such as Krugman (1979), which would imply that all trade is intra-industry.Rather, the evidence points toward hybrid frameworks such as Helpman and Krugman (1985), or Bernard et al. (2007), which allow for changes in the relative weight of sectors characterized by monopolistic competition.We will return to this in the last section of our paper.
Figures 16 and 17 provide a more nuanced image of intra-industry trade.We first show intra-industry trade by major trade partners.We see that with France or with Austria-Hungary, intra-trade was around 30 percent even at the product level, which is above the average across all trade partners.
In figure 17 we are moving from a product-level toward a variety-level perspective.The figure contains all three-digit industries in which there was more than 50 percent intraindustry trade and shows their share in total trade as of 1913.It only considers trade with Europe and North America since intra-industry trade with other continents was much less intense.Also, note that not all industries received a label to avoid clutter.
Exposing the data in this way uncovers that some industry-level trade with certain trade partners was almost entirely characterized by intra-industry trade-notwithstanding that they may have made up only a small fraction of total trade.Most of these industries were manufacturing industries (they make up most of the unlabeled dots).But this is only part of the story, because some industries with high intra-industry trade were not based in classic manufacturing: trade in hides and skins with Russia was over 80 percent intra-industry, and quantitatively non-trivial.So was yarn trade with Austria-Hungary.Firewood trade with France was even 99 percent intra-industry.This means that, while there was generally more intra-industry trade in manufacturing industries, even some non-manufacturing trade was two-way, at least with certain trade partners.Overall, it emerges that intra-industry trade was more prominent with rich economies-that is, mostly European trade partners.All this reaffirms our conjecture that models of imperfect competition (that allow, e.g., for branding or quality differences) are instrumental to explain international trade especially between European trade partners already in the first globalization.15

Changes in diversification and specialization, 1880-1913
We have seen not only that Germany's trade was growing in absolute and relative terms but also that this growth was a rather complex process with growth in new goods, trade partners, and substantial intra-industry trade.The literature on the link between growth and international trade suggests that the diversification of exports corresponds to better growth perspectives (Hausmann et al. 2007;Becuwe et al. 2018).More diversified imports, on the other hand, mean less dependence on single suppliers, which can moderate business cycle fluctuations (Calderón et al. 2007).So, how did diversification change over time?Consider a simple Herfindahl index, as with v dikt being trade in 1913 marks in direction d with partner i and SITC three-digit sector k in year t. Figure 18 plots the indices for imports and exports.Both indices fall over time suggesting growing diversification.The dashed trend lines are almost identical.But the indices suggest that exports were already in 1880 more diversified and they diversified a bit more strongly than imports over the 33 years we are looking at.As Figure 19.RCAs according to eq. 4. A: SITC One-digit sectors.B: Lall's (2000) and Rauch (1999)  suggested by Hausmann et al. (2007) this diversification was likely both the result of German economic growth and contributed to it.Quite similar to France (Becuwe et al. 2018) and Italy (Federico and Wolf 2012), Germany's foreign trade became increasingly diversified before 1914.
How can we square all this evidence on the role of the extensive margin, intra-industry trade and increasing diversification with the notion that Germany became specialized in manufacturing?In figure 7, we saw that, very broadly, Germany did export far more manufacturing products than it imported.To summarize this specialization and its changes over time, we suggest using a Lafay index (Lafay 1992;Federico and Wolf 2012).The index is based on the difference between normalized net exports for sector (or trade partner) k and the total normalized net balance weighted with the share of the product on total trade, akin to the concept of revealed comparative advantages (RCAs).Formally, with X denoting exports and M denoting imports.All k-wise indices sum up to zero and are defined between [−200, 200].Intuitively, a positive value of this index points to "success" in terms of positive net exports of k, normalized by the total trade balance and the weight of k in total trade. 16Figure 19 plots the Lafay indices with respect to four different dimensions.
The upper panels consider specialization along the product space.Figure 19A shows that Germany specialized in manufacturing products, including chemicals, machinery, and transport equipment.More detailed evidence at the three-digit level (not shown) suggests that in 1913 Germany had a strong comparative advantage not only in synthetic dies (SITC 531), fertilizers (SITC 562), railway vehicles (791), and cars (781) but also in agricultural machinery (SITC 721) or sugar (SITC 601), among many other goods.Figure 19B disaggregates comparative advantage according to the product characteristics of Rauch (1999) and Lall (2000).We see that Germany did specialize increasingly in differentiated goods, and those that embody technology (notably medium technology according to Lall, 2000, which largely reflect the technology frontier before 1914): meters and counters, non-domestic refrigerators and refrigerating equipment, refracting telescopes (something that is considered high-tech even today), electrical apparatus, radio-broadcasting transmitters, or clock and watch parts.
The data also show very clearly that it was simpler goods, where Germany de-specialized.
Finally, the lower panels of figure 19 consider the Lafay index with respect to Germany's trade partners.Figure 19C shows that Germany was increasingly successful within Europe but was in disadvantage with the rest of the world.Germany had the biggest thus defined comparative advantage (in 1913) with Great Britain and Ireland, the Netherlands, Austria-Hungary, Switzerland, France as well as Belgium and Luxembourg, among others: all rich neighbors of Germany.Germany also had a comparative advantage against Japan or the Ottoman Empire.On the other hand, Germany had comparative disadvantages with British India and the Russian Empire, that is, two of the major "emerging markets" of the time (at least from a German perspective), as well the United States and Argentina-the two largest overseas economies.
Similarly, when differentiating by income groups, figure 19D shows that German trade was successful in the rich world, particularly in the richer parts of Europe.In 1880, Germany had started with a near balanced trade result with "rich" Europe. 17 running a very large trade surplus with this group of countries.This in turn was mostly due to a trade surplus in manufacturing products, where two broad sectors made a growing contribution: chemical (sector 5) and machinery and transport equipment (sector 7).Hence, while the overall trade balance of Germany stayed negative for each year before 1914, this hides the fact that the country had become the dominant manufacturing exporter within Europe.In figure 20, we show the trade balances of particular sectors and regions 1880-1913, as a decomposition of overall trade balance shown in figure 1A.
Taken together, German trade growth over the 33 years before the World War I was largely driven by expansion along the extensive margin, with new products conquering new markets in the world.This was reflected in a substantial diversification of both exports and imports.Yet, Germany specialized with a growing trade surplus in manufacturing products, notably chemicals, machinery, and transport equipment.It is remarkable that this success was most pronounced in trade with Germany's rich neighbors within Europe-a finding that resembles the German trade patterns over the last decades.

Summary, implications and outlook
Our new data allow researchers to revisit old debates and test many new hypotheses.Let us summarize the evidence before we discuss some of its implications.During the first globalization, Germany experienced very substantial growth in exports and imports, catching up to the UK as the world's largest exporter just around 1914.We confirm the old wisdom that Germany specialized in manufacturing products, notably in chemicals, machinery, and transport equipment, and generally in differentiated products with substantial technological content.Moreover, we show how Germany established a very strong position within Europe, particularly in trade with her rich neighbors.While this is not entirely new, our data allow us to show this more systematically and in a more nuanced way than ever before (e.g., regarding differences between trade partners, technology groups, etc.).What is new, however, is our evidence that nearly all growth of Germany's trade took place along the extensive margin, with an expanding range of products and trade partners.Our evidence strongly suggests that this is not a statistical artifact but rather reflects the very nature of trade even before 1914, clearly for manufacturing but also for many non-manufacturing products.Related, we found that imports and exports grew much more diversified over time, with a substantial (and remarkably stable) share of intra-industry trade across all types of sectors and trade partners for all years 1880-1913.
Our evidence on the German Empire fits well into recent results for the pattern of trade of other European countries before 1914.Researchers have constructed new data that allow an analysis at the level of both trade partners and more or less detailed product categories (e.g., Huberman and Oosterlinck 2017, for Belgium;Federico et al. 2012 for Italy; Becuwe et al, 2018 for France).It has been shown that trade became increasingly specialized over time (Federico et al. 2012, Federico and Wolf 2012, Becuwe et al. 2018), increasingly characterized by intra-industry trade (Becuwe et al. 2021), with expanding manufacturing trade in imports and exports.Huberman and Oosterlinck (2017) show for a sample covering about 50 percent of Belgian trade that the extensive margin was important for overall trade growth.Together with our new findings for Germany, which was one of the largest economies in the world before 1914, this suggests that we need to revise our perspective on the first globalization (echoing Meissner 2015).
In order to go beyond a description of trade patterns, which is the focus of this paper, we would need a testable theoretical framework, where possible an empirical strategy to identify mechanisms, and more background about the context.On the latter, we refer the reader to the recent survey of German trade policy 1880-1914 by Hungerland and Lampe (2021).During the period covered in this paper, Germany turned from a position of free trade to more protectionism and back.Average tariff equivalents developed from around 4 percent before 1880 over 9 percent in 1900 to below 6 percent in 1913, while trade continuously grew faster than GDP.Hence, the observed trade expansion can probably not be explained by changes in trade policy alone, but likely resulted from a combination of changing technology (via transportation and productivity), institutional change (e.g., due to monetary integration), and changes in foreign demand.A systematic analysis of the effects of changes in trade costs on trade patterns would be important but is beyond the scope of our paper.
Still, our descriptive evidence has strong implications for the type of theory needed to understand the first globalization.While the standard factor proportions models in the wake of Heckscher-Ohlin can capture the broad structure of trade-such as Germany's comparative advantage in manufacturing or the pattern of trade between Germany and non-European countries-they miss nearly all of the underlying dynamics of growth and reallocation and likely also the political implications of trade liberalization.In particular, we show that the pattern of trade within Europe can hardly be explained by a factor proportions approach.Apparently, heterogeneity within even narrowly defined sectors was very high and needs to be considered.
This suggests that neither the standard narrative around Heckscher-Ohlin models nor new trade models in the wake of Eaton and Kortum (2002), Krugman (1979), or Melitz (2003) are able to explain European trade during the first globalization.Instead, it is more promising to look for theoretical frameworks that allow for within-sector heterogeneity embedded in some type of factor proportions model.Bernard et al. (2007) provide such a framework and spell out its implications.Drawing on previous work by Helpman and Krugman (1985), their model integrates the factor proportions framework of the Heckscher-Ohlin model with new trade theory assumptions about imperfect competition and economies of scale as in Melitz (2003).Such a framework can explain why some countries export more in certain sectors and industries (due to endowment driven comparative advantage) and why we nonetheless observe intra-industry trade even at the level of narrowly defined industries (due to product differentiation with increasing returns to scale).In addition, it can explain why most trade growth takes place along the extensive margin (due to within-sector heterogeneity in terms of productivity and market entry costs).Moreover, Bernard et al. (2007) show that the four wellknown theorems of the Heckscher-Ohlin model (including the Stolper-Samuelson theorem) still apply, but with some important qualifications.For example, the interaction between within-sector heterogeneity and comparative advantage gives rise to additional welfare gains from trade.Among other things, such gains will limit the decline of the real price of scarce factors and hence mitigate the distributional effect of trade liberalization according to the Stolper-Samuelson theorem.Depending on parameters and endowment differences, the real reward of scarce factors can even increase in response to declining trade costs.While our data start in 1880, which is after the major protectionist turn of 1879, it would be interesting to revisit the political economy of trade policy before 1914 as discussed by Klug (2001) or Lehmann (2010).This is related to the idea of "smooth adjustment" to trade shocks as highlighted by Becuwe et al. (2021).All this clearly invites new empirical work.For example, it would be possible to combine our disaggregated data for Germany with data on France (Becuwe et al. 2018), Italy (Federico et al. 2012), Belgium (Huberman and Oosterlinck 2017), and others, using the SITC classification scheme to bridge historical statistical systems.Such a combined dataset would allow to cross-check historical sources and lend itself to many empirical applications.
Finally, our evidence on heterogeneity within sectors and industries points to the role of firms as major actors in international trade already during the first globalization.While this has long been recognized by business historians and documented in excellent case studies for specific sectors (e.g., Brown 1992;Brown 1995), it was difficult to reconcile with the big picture along the lines of Heckscher-Ohlin.Unfortunately, we lack systematic firm level data on trade, but our findings on the importance of the extensive margin suggest searching for related firm-level evidence, such as firm sizes (see, e.g., Biermann 2021).High-productivity firms probably benefitted much more from declining trade costs, notably in sectors with a comparative advantage as suggested by Bernard et al. (2007).This opens a new perspective on the rise of German multinationals before 1914 such as Bosch, Siemens, or BASF to name a few.

Figure 1 .
Figure 1.A: In 1913-marks.Statistical items excluded.B: Value of German exports and imports in percent of trade with the UK .Source: Own calculations and Federico-Tena World Trade Historical Database.

Figure 2 .
Figure 2. Number of active trade partners.Source: Own calculations.

Figure 3 .
Figure 3. A: Trade (imports plus exports) per continent in percent of total trade.B: Trade balance (imports minus exports, pooled over entire period) in million 1913-marks per continent.Source: Own calculations.

Figure 4 .
Figure 4. Germany's imports in 1913 by trade partner.Shares in traded value.Source: Own calculations.

Figure 5 .
Figure 5. Germany's exports in 1913 by trade partner.Shares in traded value.Source: Own calculations.

Figure 6 .
Figure 6.A:Number of products.B:Average SITC item length.Source: Own calculations.

Figure 7 .
Figure 7. A: Imports.B: Exports.In percent of total imports.Source: Own calculations.

Figure 10 .
Figure 10.Available and active varieties.Source: Own calculations.

Figure 13 .
Figure 13.Margins, 1880-1913.A: Europe.B: Rest of the world.In percent of total trade according to eq. 1.Source: Own calculations.

Figure 16 .
Figure16.Intra-industry trade with major trade partners at the product-level (SITC fivedigit).In percent of total trade with respective trade partner according to eq. 2. Source: Own calculations.

Figure 17 .
Figure 17.Sectors with high shares of intra-industry trade.A: Europe.B: North America.In percent of total trade according to eq. 2. Dots reflect trade within an SITC three-digit industry in 1913.Source: Own calculations.

Figure 18 .
Figure 18.Diversification.Herfindahl indices according to eq. 3 based on country-sector pairs within SITC three-digit sectors.Dashed lines = trend line based on linear fit.Source: Own calculations.
Figure 20.Trade balances in million 1913-marks.A: Broad sectors.B: Main manufacturing sectors.C: Europe vs. rest of the world.D: Main manufacturing sectors in percent of total manufacturing trade.Sources: Own calculations.

Table 1 .
Comparison of trade between 1880 and 1888 by regionValues in million 1913-marks.'NA' denotes accounting items (Sonderposten) with no geographical direction like 'ships' supplies' or the like.See table A1 in the Appendix for a classification of countries into Western and Eastern Europe.Source: own calculations.

Table 2 .
Descriptive statistics on the number of countries per year From 1880 to 1913.Statistical items excluded.Source: Own calculations.

Table 3 .
Sectoral trade growth

Table 4 .
Available and active varieties, 1880 and 1913 At SITC five-digit level.Source: Own calculations.
classifications.C: Continents.D: Income classification; see table A2 in the Appendix for a classification of countries.Source: Own calculations.